National estimates of the impact of electronic health records on the workload of primary care physicians

We used the 2006–2009 National Ambulatory Medical Care Survey (NAMCS). The NAMCS is a cross-sectional national probability sample survey administrated by the National Center for Health Statistics (NCHS) for the Centers for Disease Control and Prevention (CDC) [21]. The NAMCS collects data on patient office visits to non-federally employed office-based physicians in the United States. For each visit, physicians or staff members complete a one page survey form containing patient demographics, reasons for the visit, physicians’ diagnoses, and physicians’ time spent in face-to-face patient care. Detailed descriptions of the survey are available from the authors and on the NAMCS website (http://​www.​cdc.​gov/​nchs/​ahcd.​htm). We examined 37,962 patient visits to 1470 primary care physicians during 2006–2009 (general practitioners, family practitioners, and general internists). Through the NCHS Research Data Center, we have obtained access to a restricted version of the NAMCS data with information on physician’s age and the number of patient visits per physician per week. We use 2006–2009 since it is the baseline period prior to the introduction of the HITECH EHR adoption subsidies.

To gauge productivity/efficiency gains in primary care physician workloads, we use (1) duration of time measured in minutes of the face-to-face encounter between physicians and patients (patient face time) for direct patient care during the office visit, and (2) number of total patient office visits per physician per week (patient volume). Patient face time, self-reported by physicians, only includes time physicians spent with their patients. Time spent waiting to see physicians, reviewing patients’ medical records by physicians before seeing patients, and receiving care from non-physician clinicians are excluded. Physician’s number of patient visits per week is the count of patient visits during the survey week reported by physicians or office staffs.

The key independent variable of this study is a binary EHR adoption variable “basic EHRs” defined for NAMCS data by DesRoches et al. [22] The basic EHRs include at least 5 fundamental EHR functions: (1) patient demographic information, (2) clinical notes, (3) computerized orders for prescriptions, (4) viewing laboratory results, and (5) viewing image results.

A cross-sectional analysis of pooled survey data was conducted. Multivariate regression techniques were used to estimate the impact of the basic EHR use on the following outcomes of interest: patient face time per visit and physician’s number of patient visits per week. Ordinary least square (OLS) regressions will only capture the effects of EHR at the mean visit time and mean patient volume. But, the effect of EHR on face time may vary depending on whether the visit is particularly short or long, and the effect on patient volume may vary depending on whether the physician has a small or large practice. To capture this variation in the effects of EHR, a novel feature of this paper is that we use simultaneous-quantile regressions (SQR) in Stata 12 [23]. In our visit-level face time regressions, SQR simultaneously estimates the impact of EHR on face time at five points in the distribution of face time: 10 min (12^th percentile), 15 min (15^th percentile), 20 min (60^th percentile), 25 min (75^th percentile), and 30 min (80^th percentile). Standard errors in SQR are bootstrapped. In the physician-level patient volume regressions, SQR simultaneously estimates the impact of EHR on volume at three points in the distribution of patient volume: the 10^th, 50^th, and 90^th quartiles of patient volume (20, 64, 120 visits per week). Time and volume are logged in all regressions.

Next, in each regression we construct indicators for young physicians (age 36 or younger) (12^th percentile) and older physicians (age 62 or older) (88^th percentile) and include interaction terms between these two indicators and ‘basic EHR’ to examine how the effect of the basic EHR use on physicians’ workload productivity varies across physician’s age. In addition to these interactions, our regressions control for 46 covariates: patient characteristics, health status of patients, visit characteristics, physicians characteristics, physician group characteristics, and other covariates such as geographic region, metropolitan statistical area (MSA) status, and survey year. MSA is delineated by the US Office of Management and Budget Census as a “core urban area of 50,000 or more population." MSA has been commonly used in literature to define urban/rural status. All the covariates with descriptive statistics are reported in Appendix 1.

During 2006–2009, 19.9 % of primary care physicians had basic EHRs. Older physicians were less likely to adopt basic EHRs than younger physicians (column 1, Table 1). While 23.1 % of physicians aged 36 or less and 20.4 % of physicians aged between 37 and 61 had basic EHRs, only 13.6 % of physicians aged 62 or older had basic EHRs. Physicians with basic EHRs spent longer face time with their patients than physicians without basic EHRs across all physician age groups (column 2, Table 1). In general, under EHR adoption, physicians had fewer patient visits per week than without EHR. However, this was reversed for older physicians, who had more patient visits with EHR than without (column 3, Table 1). Moreover, older physicians that adopted EHR had more visits per week than the young and middle-aged physicians that adopted. In the case of non-EHR adopting physicians, older physicians had fewer visits per week than the young and middle-aged physicians. Thus, the raw data indicates that EHRs are associated with increases in the physician’s face time with the patient regardless of physician age, whereas the association between EHRs and patient volume varies across physician age. Next, we see if these patterns persist once we control for patient and physician characteristics in multivariate regressions.

Figure 1 presents the impact of basic EHR on patient face time at selected levels of patient face time (10, 15, 20, 25, 30 min) by physician age group, estimated by quantile regressions controlling for 46 covariates (the regressions are detailed in Appendix 2). Figure 1 is presenting the full incremental effects of EHR estimated from Table 2.¹ In Fig. 1, young physicians (age 36 or younger) using basic EHR consistently increased patient face time by 7-9 % at all points in the time distribution (p < .01). On the contrary, older physicians did not increase patient face time for visits of 10 min and 15 min (statistically insignificant), but raised patient face time by 12 %, 14 %, and 15 % for patient visits of 20 min, 25 min, and 30 min, respectively (p < .01).

Overall, in terms of absolute magnitude, these effects can be expressed as follows: for 15 min visits, older physicians spent 1.4 min less per visit than younger physicians due to EHR, but for 30 min visits, they spent 1.8 min more than younger physicians due to EHR. Thus, compared to younger physicians, there is a 3.2 min increase in face time over the range of visits (15 to 30 min) for older physicians that is entirely due to EHR. Note that at the mean visit time (19 min), young and old physicians behave about the same, spending 7 % more time with the patient (1.3 min, p < 0.001) under EHR than without EHR.² These extra 1.3 face time minutes per visit result in a total of 1.5 extra hours of face time per week under EHR. In the subsequent section, we will examine whether the increase in patient face time causes physicians to reduce the number of visits over the week.

Figure 2 shows the effect of basic EHR use on the physician’s number of patient visits per week at the 10^th, 50^th, and 90^th quantiles of patient volume (20, 64, 120 visits per week) by physician age group, estimated by quantile regressions (see Table 3). While basic EHR use reduced the young physician’s number of patient visits by 17-22 %, older physicians increased number of visits by 30-40 %, consistently across all scales of practice sizes. This indicates that EHRs are associated with increased weekly patient volume among older physicians and with reduced patient volume among young physicians. In column 1, Table 3, the overall average impact of EHR on the young is the sum of the coefficients: EHR + EHR*young = -.027-.185 = -.212. Not shown, the standard error for this sum is estimated to be .126, so the full effect of -0.212 has a p-value < 0.09. Log retransformed, this is an effect of size -.19 as the average effect of EHR on the number of weekly visits for young physicians. For older physicians, the average effect is + .44 (p = 0.015). This amounts to an increase of 26 visits per week for older physicians with EHR compared to older physicians without EHR. For younger physicians, EHR is associated with a reduction of about 12 visits per week at the mean. Overall, young physicians with EHR have 53 visits per week compared to 85 visits for older physicians with EHR. When there is no EHR adoption, young physicians have 65 visits per week compared to 59 visits for older physicians. Among young physicians, EHR use is associated with a decline in patient volume, while EHR use among older physicians is associated with an increase in volume. However, the countervailing effects between younger and older physicians cancel each other out, so that on average, basic EHR use had no statistically significant impact on overall physician visit volume per week at all levels of the volume distribution.